Efficient Behavior Selection for Immunologically-inspired Agents

This paper describes and evaluates a biologically-inspired adaptation mechanism that allows network application components (agents) to autonomously adapt to dynamic environment changes in the network (e.g. changes in network traffic and resource availability). Based on the observation that the immune system elegantly achieves autonomous adaptation, the proposed mechanism, called the iNet artificial immune system, is designed after a scheme in which the immune system produces specific antibodies against an antigen invasion. iNet models an agent behavior (e.g. migration and replication) as an antibody, and an environment condition (e.g. network traffic and resource availability) as an antigen. iNet allows each agent to (1) autonomously monitor its surrounding environment conditions (i.e. antigens) to evaluate whether it adapts to the environment conditions, and if it does not, (2) adaptively perform a behavior (i.e. antibody) suitable for the environment conditions. This paper focuses on the first process, the environment evaluation process, and describes its design and implementation. The experimental results show that the environment evaluation process works efficiently with a high degree of accuracy.